Working Paper
Series
WP 13-3
MAY 2013
Does High Home-Ownership Impair the
Labor Market?
David G. Blanchflower and Andrew J. Oswald
Abstract
We explore the hypothesis that high home-ownership damages the labor market. Our results are relevant to, and may be
worrying for, a range of policymakers and researchers. We find that rises in the home-ownership rate in a US state are
a precursor to eventual sharp rises in unemployment in that state. The elasticity exceeds unity: A doubling of the rate
of home-ownership in a US state is followed in the long-run by more than a doubling of the later unemployment rate.
What mechanism might explain this? We show that rises in home-ownership lead to three problems: (i) lower levels of
labor mobility, (ii) greater commuting times, and (iii) fewer new businesses. Our argument is not that owners themselves
are disproportionately unemployed. The evidence suggests, instead, that the housing market can produce negative ‘externalities’ upon the labor market. The time lags are long. That gradualness may explain why these important patterns are so
little-known.
JEL codes: I1, I3.
Keywords: Natural rate of unemployment; labor market; housing market; structural; business cycles; mobility.
David G. Blanchflower is the Bruce V. Rauner ‘78 Professor of Economics at Dartmouth College, a part-time professor
at the University of Stirling; a Research Associate at the NBER and a Nonresident Senior Fellow at the Peterson Institute.
He was a member of the Monetary Policy Committee at the Bank of England from 2006–09. Andrew J. Oswald is
a professor of Economics at the University of Warwick in England and a Research Associate at the ESRC Centre for
Competitive Advantage in the Global Economy (CAGE). He has been a visiting professor at Cornell, Dartmouth,
Harvard, and Princeton.
Note: For suggestions and valuable discussion, we thank seminar participants at PIIE and Hank Farber, William Fischel,
Barry McCormick, Ian M. McDonald, Michael McMahon, Robin Naylor, Jeremy Smith, Doug Staiger, Rob Valletta,
and Thijs van Rens.
Copyright © 2013 by the Peterson Institute for International Economics. All rights reserved. No part of
this working paper may be reproduced or utilized in any form or by any means, electronic or mechanical, including
photocopying, recording, or by information storage or retrieval system, without permission from the Institute.
1750 Massachusetts Avenue, NW Washington, DC 20036-1903
Tel: (202) 328-9000 Fax: (202) 659-3225 www.piie.com
INTRODUCTION
Unemployment is a major source of unhappiness, mental ill-health, and lost income.1 Yet after a century
of economic research on the topic, the determinants of the equilibrium or ‘natural’ rate of unemployment
are still imperfectly understood, and unemployment levels in the industrialized nations are today 10
percent, with some nations over 20 percent.2 The historical focus of the research literature has been on
which labor-market characteristics—trade unionism, unemployment benefits, job protection, etc.—are
particularly influential.
We propose a different approach to the problem. Our study provides evidence consistent with the
view that the housing market plays a fundamental role as a determinant of the rate of unemployment.
We study modern and historic data from the United States. We construct state panels, and then estimate
unemployment equations.3 Using data on two million randomly sampled Americans, we also estimate
equations for the number of weeks worked, the extent of labor mobility, the length of commuting times,
and the number of businesses. Our results seem relevant to, and should perhaps be seen as worrying for,
a wide range of policy-makers and researchers. There are four main conclusions. First, we document a
strong statistical link between high levels of home-ownership in a geographical area and later high levels
of joblessness in that area. We show that this result is robust across sub-periods going back to the 1980s.
The lags from ownership levels to unemployment levels are long. They can take up to five years to be
evident. This suggests that high home-ownership may gradually interfere with the efficient functioning
of a labor market. Second, we show that, both within states and across states, high home-ownership areas
have lower labor mobility. Importantly, this is not due merely to the personal characteristics of owners and
renters. We are unable, in this paper, to say exactly why, or to give a complete explanation for the patterns
that are found, but our study’s results are consistent with the unusual idea that the housing market can
create dampening externalities upon the labor market and the economy. Third, we show that states with
higher rates of home-ownership have longer commute times. This phenomenon is likely to be a reflection
of the greater transport congestion that goes with a less mobile workforce and it will act to raise costs for
employers and employees. Fourth, we demonstrate that states with higher rates of home-ownership have
1. Linn, Sandifer, and Stein (1985), DiTella, MacCulloch, and Oswald (2003), Murphy and Athanasou (1999), Paul and
Moser (2009), and Powdthavee (2010), for example.
2. The euro area unemployment rate for December 2012 was 11.7 percent, ranging from a low of 4.3 percent in Austria,
5.3 percent in Germany, and 5.8 percent in the Netherlands, up to a high of 26.8 percent in Greece and 26.1 percent in
Spain, both of which had youth unemployment rates of greater than 50 percent. France had a rate of 10.6 percent and
Italy 10.9 percent, while the United Kingdom and the United States both had unemployment rates of 7.8 percent. See
http://epp.eurostat.ec.europa.eu/cache/ITY_PUBLIC/3-01022013-BP/EN/3-01022013-BP-EN.PDF.
3. Our work builds upon a tradition of labor-market research with state panels from the 1990s in sources such as
Blanchard and Katz (1992) and Blanchflower and Oswald (1994).
2
lower rates of business formation. This might be the result of zoning restrictions and NIMBY effects of a
kind that, as Fischel (2004) discusses, are rational for homeowners. But currently that channel can be only
a conjecture.
The data used in this paper are almost wholly from the United States. However, our conclusions
may have wider implications. Taken in conjunction with new work by Laamanen (2013), which was
done independently of our own,4 and reaches similar conclusions for the country of Finland, the findings
may go some way to explain why nations like Spain (80 percent owners, 20+ percent unemployment)
and Switzerland (30 percent owners, 3 percent unemployment) can have such different mixtures of
home-ownership and joblessness. Figure 1 shows that there is a strong positive correlation across the
wealthy countries between their home-ownership rates and the latest unemployment rates. Such a figure
is open to the sensible criticism that the scatter might be an illusion caused by country fixed-effects.
However, that objection cannot be raised about figure 2, which is for the United States. It plots very long
changes (approximately half-century changes) in home-ownership rates and unemployment rates across
the US states—minus Alaska and Hawaii—and generates a similar result. It plots the 50-year change in
home-ownership rates (1950–2000) against a 60-year change in unemployment rates (1950–2010).5
The later analysis does not depend on data from the special period of the 2007 US house-price
crash;6 nor does it rely on the idea that homeowners are themselves disproportionately unemployed (there
is a considerable literature that suggests such a claim is false, or, at best, weak); nor does it imply that
spatial compensating differentials theory is incorrect; nor is it Keynesian in spirit;7 nor does it rest upon
the idea of house-lock in a housing downturn (for example, Ferreira, Gyourko, and Tracy 2010, Valletta
2012). Our paper makes a simple statistical contribution and discusses possible mechanisms. The detailed
nature of any housing-labor externality remains poorly understood.
BACKGROUND
In his 1968 address to the American Economic Association, Milton Friedman famously argued that
the natural rate of unemployment can be expected to depend upon the degree of labor mobility in the
economy. The functioning of the labor market will thus be shaped not just by long-studied factors such
4. In April 2013, Laanamen and Blanchflower and Oswald discovered they had equivalent empirical findings, though done
in different ways, for Finland and the United States respectively.
5. Source for the 1950 state unemployment rates is http://www.census.gov/prod/www/abs/decennial/1950cenpopv2.html
6. Repercussions from the worldwide house-price bubble are discussed in sources such as Bell and Blanchflower (2010)
and Dickens and Triest (2012).
7. We would like to acknowledge valuable discussions with Ian McDonald on this issue. One reason why our effect does
not appear to be consistent with a Keynesian argument is that we find the lags from home-ownership are long, and that is
inconsistent with the idea that our estimated unemployment effect in time t is the result of aggregate demand in time t.
3
as the generosity of unemployment benefits and the strength of trade unions,8 but also by the nature, and
inherent flexibility and dynamism of, the housing market. However, on that topic there has been relatively
little empirical research.
One important early line of work stemmed from scholars such as McCormick (1983) and
Hughes and McCormick (1981). This found evidence that in certain types of public-sector housing the
degree of labor mobility was low and the associated joblessness was high.9 That research tradition still
continues—as in Dujardin and Goffette-Nagot (2009). A broader literature at the border between labor
and urban economics has considered whether there might be fundamental differences in the labor-market
impact of renting rather than owning. Some of this work was triggered by the suggestion in a public
lecture by Oswald (1996, 1997) that, especially in Europe, at the aggregate level a higher proportion
of home-ownership (or ‘owner-occupation’) seems to be associated empirically with a larger amount of
unemployment. Oswald’s data were mainly for the western nations and for the US states. He presented
no formal regression equations. Green and Hendershott (2001) subsequently reported US econometric
results that were somewhat, though not entirely, supportive.
One theoretical interpretation of these early patterns was that home-ownership might raise
unemployment by slowing the ability of jobless owners to move to new opportunities. In response to
this idea, a number of researchers later examined micro data. The ensuing literature concluded that the
bulk of the evidence is against the idea that home-owning individuals are unemployed more than renters.
Hence—though the empirical debate continues—a number of authors concluded that Oswald’s general
idea must be incorrect and the cross-country pattern must be illusory. A modern literature includes Battu,
Ma, and Phimister (2008), Coulson and Fisher (2002, 2009), Dohmen (2005), Head and Lloyd-Ellis
(2012), Van Leuvensteijn and Koning (2004), Munch, Rosholm, and Svarer (2006), Rouwendal and
Nijkamp (2010), Smith and Zenou (2003), and Zabel (2012).
An alternative possibility—one that has not, to our knowledge, been fully explored empirically—is
the hypothesis that the housing market might create externalities. There are a number of ways in which
such spillovers might operate. For example, Serafinelli (2012) shows in the US labor market that there
appear to be beneficial informational externalities upon workers’ productivity from a high degree of
labor mobility. Although the author does not pursue the implication, this raises the possibility that
any housing market structure that led to immobility could, therefore, produce negative externalities
on workers and firms. Oswald (1999) suggests a different possible channel. Homeowners might act to
hold back development in their area (through zoning restrictions) in a way that could be detrimental to
8. For example, OECD (1994) and Layard, Nickell, and Jackman (1991).
9. McCormick (1983) makes the interesting point that economists should not work on the assumption that low mobility
is always undesirable. If home-ownership facilitates the accumulation of wealth, and wealth has a negative effect on
migration rates, then a migration cost arises which will and should influence migration and hence unemployment rates,
without necessarily doing “bad things.”
4
new jobs and entrepreneurial ventures. This would be NIMBY pressures in action. A third possibility
is that regions with high home-ownership might be difficult ones in which to attract migrant workers
(who may require the flexibility of rental accommodation). As a fourth possibility, a formal model in the
literature by Dohmen (2005) predicts that high ownership can be associated with high unemployment.
The reason, within Dohmen’s framework, is not one linked to an externality but instead to the fact that
the composition of the unemployed pool is endogenous to the structure of the housing market (in other
words, the kind of person who is unemployed alters when the home-ownership rate goes up). None of
these mechanisms requires the homeowners themselves to be disproportionately unemployed (as in the
critique of Munch, Rosholm, and Svarer 2006).
Most unemployment researchers begin from the tradition of neoclassical economics and with the
idea that there is some underlying equation, defined over tastes and technology, that explains the structural
or long-run rates of unemployment and employment. Whether from the modern matching tradition due
especially to researchers such as Mortensen and Pissarides (1994), the 1990s macro-labor literature due
especially to researchers such as Layard, Nickell and Jackman (1991), or the classical literature that goes
back at least to Pigou (1914), a huge body of empirical work in economics has searched for labor market
characteristics—such as the degree of trade unionism—that might enter that natural-rate equation.
We wish to remain open-minded about the true model of the labor market. For a region’s
unemployment rate, we will think about this generically as an autoregressive relationship that has a steady
state solution, U*, which we will think of as the natural rate of unemployment. In estimation, we can
view the relevant equation as:
Unemployment rate in a region U(t) = f(U(t-1), labor market characteristics, housing market
characteristics, people’s demographic and educational characteristics, region characteristics, year dummies)
We will add the rate of home-ownership in an area to the usual list of variables. For some nations, it
would be ideal to allow for a division of the housing market into three broad segments—owners, private
renters, and public-sector renters. In our empirical work, however, we use data from the United States,
where public renting is sufficiently rare that it can be largely neglected.
EMPIRICS
Tables 1 and 2 document the raw data. As implied by the earlier figure 2, home-ownership in the United
States has grown strongly since 1900. It changed from a mean of approximately 46 percent in that year
to approximately 65 percent by the year 2010. Table 2 shows that the US rate peaked in the year 2004,
at 69 percent. This was a few years before the start of the infamous modern housing crash. In 2010, the
US states with the highest levels of home-ownership were states such as Minnesota, Michigan, Delaware,
5
and Iowa. The ownership levels were lowest in states such as California and New York (and in DC if
viewed as a state).10
In tables 3 and 4, we estimate unemployment equations. The estimation here is on an annual panel
of US states and uses as its dependent variable the natural logarithm of the state unemployment rate. Our
data cover a quarter of a century of consecutive years and are drawn from the Merged Outgoing Rotation
Groups of the Current Population Survey. The exact period is 1985 to 2011, which gives us an effective
sample size of 1377 area-time observations (that is, of states by years). The different columns of tables 3
and 4 lay out a range of specifications, including autoregressive specifications with a lagged dependent
variable. Table 3 is presented for intellectual completeness rather than because we believe it to be an
adequate specification (a lagged dependent variable enters strongly significantly, as would be expected, so
table 3 should not be used to draw reliable inferences).
What emerges most clearly from tables 3 and 4 is a correlation between unemployment in a state
in time t and the rate of home-ownership in that state in year t-3 and earlier. Summarizing, we find that
unemployment:
 Is higher in states that had high home-ownership rates in the past. The long-run elasticity varies from
0.8 to 1.5. Given the context, these are large numbers.
 Is autoregressive, with a coefficient on the lagged dependent variable Ut-1 of approximately 0.8. Hence
long-run effects are greatly magnified compared to the short-run or ‘impact’ effect of the independent
variable.
 Is uncorrelated with union density in the state. The appendix shows that that is also true of
unemployment insurance (UI) generosity.
 Is correlated, as would be expected, with the personal characteristics of workers in the state (the
detailed results are not reported but follow the usual pattern of joblessness being greater among those
with fewer qualifications).
 Is not significantly correlated with current home-ownership (the detailed results not reported), which
is why tables 3 and 4 begin, in their respective column 1s, with an ownership variable in time period
t-1.
10. The most recent quarterly data on home-ownership rates suggest that the decline may have slowed (percent).
Q1
Q2
Q3
Q4
Q1
Q2
Q3
Q4
2012
65.4
65.5
65.5
65.4
2011
66.4
65.9
66.3
66.0
2008
67.8
68.1
67.9
67.5
2007
68.4
68.2
68.2
67.8
2010
67.1
66.9
66.9
2009
67.3
67.4
67.6
66.5
2006
68.5
68.7
69.0
68.9
67.2
2005
69.1
68.6
68.8
69.0
Available at http://www.census.gov/housing/hvs/files/qtr412/q412press.pdf.
6
These judgments are from pooled cross-sections, so they describe associations in the data. We should
be cautious before imputing meaning into such patterns. Nevertheless, the fact that the key correlation
exists with a housing market variable so heavily lagged (back to t-5), and that the regression equations
control for state and year effects, suggests that the pattern is of interest and deserves to be taken seriously.
Figure 3 plots the impulse-response function for one algebraic example.
Table 3 contains the simplest results. In column 1, the home-ownership variable enters negatively
with a coefficient of -0.3282 and a t-statistic larger than two. If the underlying economic prosperity in a
state is captured in that year by data on either its current (high) home-ownership rate, or by data on its
(low) unemployment rate, a negative correlation here is not surprising. Columns 2 and 3 of table 3 then
examine longer lags, and the ownership variable becomes insignificantly different from zero.
Columns 4, 5, and 6 of table 3 contain results that are more remarkable. We see here that the lagged
home-ownership rate in the state becomes a powerful positive predictor of later unemployment in that
state. This finding is consistent with—though it does not prove—the claim that, after some years have
passed, a high degree of owner-occupation in an area can have a deleterious structural effect on the labor
market in that area. Table 3 reveals that the coefficient on home-ownership in an unemployment equation
becomes larger as the time lag becomes longer. In column 5 of table 3, the long-run home-ownership
elasticity of unemployment is 0.8. These equations control for a number of independent influences on the
rate of joblessness. Importantly, both state and year dummies are included in the estimation.
Table 3 is useful for illustrative purposes but is not a natural or believable specification. It is known
that, presumably because unemployment is a stock variable and not a flow variable, unemployment
equations empirically are typically highly auto-regressive and thus (because stocks are best characterized
by differential equations) need to include a lagged dependent variable. We now turn to that kind of
specification.
Table 4 contains the paper’s most important results. In the first column of table 4, for the
entire period up to 2011, a lagged dependent variable has a coefficient of 0.8482 (with a t-statistic of
approximately 50). Column 1 of table 4 includes a set of year dummies; a set of state dummies; 18
dummy variables for different levels, in the underlying micro data, of people’s education; and controls for
personal characteristics such as the average age of people in the state. The unemployment rate in this form
of panel is a slow-adjusting variable, and that holds true despite the inclusion of state fixed effects. In this
first column of table 4 the coefficient on lagged home-ownership is 0.2488. Here the lag is only a single
year. The t-statistic on this coefficient is 2.73, so the null hypothesis of a zero coefficient can be rejected
at conventional levels of confidence. The coefficient on union density has the wrong sign to be a signal of
11
any deleterious effect on joblessness; it is negative, with a t-statistic of only 0.71. These estimates allow
for adjustment for clustered standard errors.
11. In the appendix we also examine the impact of state unemployment benefits but can find no effect.
7
Because this regression equation in column 1 of table 4 is effectively a first-order difference
equation, the long-run home-ownership elasticity of unemployment is larger than the impact effect of
approximately 0.25. The long-run effect is, more precisely, 0.2488 divided by (1.0000 – 0.8482). Hence
the long run, or steady state, elasticity is estimated here at 1.7. In this context, that is a surprisingly large
number, and suggests the possibility that there are profound connections between the workings of the US
housing market and its labor market.
Interestingly, the size of the coefficient strengthens as we go further back. Column 2 of table 4
introduces a further lag on the home-ownership rate variable, namely, for ownership in year t-2. It enters
with a coefficient of 0.3359. The null hypothesis of zero can again be rejected; the t-statistic is 3.69.
In columns 3, 4, and 5, respectively, further and further lags on home-ownership are included. In
the fifth column of table 4, for example, the lagged dependent variable has a coefficient of 0.7840 and a
coefficient on home-ownership in t-5 of 0.4302. The implied long run elasticity is approximately 2.
The final column of table 4 gives the fullest kind of specification where all home-ownership rates are
included from t-1 to t-5. The sum of these coefficients is approximately 0.49. The long run relationship
thus continues to be a large one—in this case with a steady state elasticity of 2.2.
Is the pattern robust? Our experiments suggest that it is. First, it is conceivable that unemployment
and home-ownership simply both follow a state-level business cycle but with different lagged timing.
One way to probe for this is to replace the state and year dummies with state time trends; it turns out
that the results are then essentially unchanged (results available on request).12 Splitting the data into
two sub-periods, as in table 5, provides another illustration. It reveals the apparently reliability of the
correlation between the log of unemployment in period t and the log of home-ownership in a much
earlier year. In each of the two sub-periods in table 5, lagged home-ownership enters significantly. For
the period 1989–2001, the first segment of table 5 estimates the coefficient on home-ownership in t-5 at
0.3566. The coefficient on the lagged dependent variable in this estimated equation is 0.7169. Therefore
the long-run home-ownership elasticity of unemployment is 1.3. In the later period, the coefficient on
ownership in t-5 is 0.6246, and the lagged dependent variable’s coefficient is 0.6844. Then the long-run
elasticity is 2.0.
Some economists may prefer to focus on the level of employment as a key variable rather than
on the rate of joblessness itself. For that reason, table 6 replicates the same general finding using
12. Another possibility, suggested to us by Barry McCormick, is that both unemployment and home-ownership are driven
by a common state level business cycle with different lag structures. Our correlation might then be illusory. We tested for
this by estimating a series of state level home-ownership equations, which included long lags on both the log of the homeownership rate (5 lags) and the log of the unemployment rate (7 lags). There was no evidence of any effect from long
lagged unemployment rates, which suggests that home-ownership here is not driven by local business cycles.
8
employment-rate, rather than unemployment-rate, data. Lagged home-ownership rates enter strongly
negatively in this state panel equation.
Table 7 tries a different check and turns to micro data on individual workers. It estimates a weeksworked equation using data from the March Current Population Surveys between 1992 and 2011. The
sample size is approximately 2 million individuals. The dependent variable in table 7 is the number of
weeks an individual worked during the previous year, rather than the more usual ‘point of time’ measure
of whether an individual was unemployed on the day they were surveyed. Their answers are reported in
8 size bands in the data set; we allocated mid-points. Non-workers were allocated zero weeks worked. In
table 7 we include controls for year and state, as well as personal controls for race, gender, and education,
along with whether the individual was a mover—defined as whether they changed their place of residence
over the year as well as if they primarily worked full-time and/or were a union member. In addition, we
include controls for whether the individual was a homeowner or a renter (with the excluded category
being renters who received the accommodation for no charge). We also include the log of the state*year
home-ownership rate, defined at the household level, lagged four years in columns 1 and 3 and five years
in columns 2 and 4. Separate results are presented for the full sample as well as for the (vast) majority
who were non-movers. Consistent with earlier results, lagged home-ownership enters negatively with a
large coefficient in table 7 (though it hovers close to the 95 percent cut-off level of confidence). It is thus
associated with fewer weeks worked. This is equivalent to our earlier finding: It seems to imply that the
‘natural rate of employment’ is reduced by high ownership of homes. What is notable about table 7 is that
we are apparently picking up deleterious effects on the labor market after controlling for the individual
worker’s own housing status (that is, whether he or she is an owner).
INTERPRETING THE PATTERNS
Many labor economists who look at these equations will wonder about a possible role for, and any consequences of the housing market’s structure upon, the degree of labor mobility in the United States. We
turn, therefore, to tables 8 to 11.
Using US Census data, table 8 reports for a long run of years—with some gaps because of missing
data—the mean values for a variety of different measures of mobility. Five columns are given. The data in
the first column are on the total proportion of citizens moving their place of residence during the year. In
2010–11, for example, 11.6 percent of Americans moved home. In 1947–48, the figure was 20.2 percent.
This much larger number was not because of the closeness of 1947 to the end of the Second World War:
Until the end of the 1960s the proportion of movers continued to be approximately 20 percent. As
column 1 of table 8 shows, from the 1970s to the 2000s there is evidence of a secular downtrend in the
percent movers category. Even before the housing crash of 2008–09, the percentage of Americans moving
residence had fallen to approximately 14 percent per annum.
9
In columns 2 to 5 of table 8, the figure for total percentage movers is disaggregated. For 2010–11,
for example, the total figure of 11.6 percent was made up of:
11.6 percent total movers = 7.7 percent movers within the county + 2.0 percent movers within the state +
1.6 percent movers out of state to another state + 0.4 percent movers out of the United States itself
This breakdown gives an arithmetical sense of the huge degree of geographical flows that go on within
a 12-month period. As the identity equation in italics makes clear, there are different ways to measure
‘mobility’ in the United States. We start analytically with the left-hand side variable, namely, percent total
movers.
To get a sense of the relationship between the rate of geographical movement and the
home-ownership rate, table 9 provides a set of micro-econometric equations. In this case the dependent
variable is a zero-one variable for whether the survey respondent moved home in the preceding 12
months. In the first column of table 9, state and year dummies are included, but the only other
independent variables are the state unemployment rate and the home-ownership rate in the state. As
might be expected from standard economic theory, people are more likely to have moved if there is a
high degree of joblessness in their state (the coefficient is 0.0127 with a t-statistic of 3.44). Moreover,
areas with a high level of home-ownership have lower mobility, ceteris paribus. The coefficient in the first
column of table 9 is −0.1766 with a large t-statistic. This implies that the rate of movement in a state is
nearly 18 percentage points lower in a place with double the home-ownership rate of another area.
The coefficients in column 1 of table 9 are only marginally influenced by the addition to the
regression equation of a set of personal controls (such as the individual’s age and level of education).
Columns 2 and 3 demonstrate the apparent robustness. The coefficient on the logarithm of state
home-ownership falls in size only slightly to −0.1542. However, as more variables, and in particular
ones for whether the individual is a renter or owner, are added to the columns of table 9, the log of the
home-ownership rate in the state drops. If standard errors are clustered at the state level alone, then the
variable loses statistical significance. As might be expected, the variables for the literal housing status
of the person are strong. Being a renter is associated with a much greater chance of being a mover. In
the final column of table 9, that coefficient is 0.2134. In this equation, the coefficient on the log of
home-ownership is −0.0633 with a t-statistic of 1.62.
Because mobility can be conceived of more broadly, table 10 estimates equations instead for both
within-state and out-of-state mobility combined. In this case there is evidence of a robust negative effect
of state home-ownership upon the rate of mobility. The first column of table 10 estimates the state-panel
equation:
10
Log Mover Rate in t = f(Log Mover Rate in t-1, Log Unem Rate in t-1, Log Union Density Rate in t-1,
Log Home-ownership Rate in t-1, state dummies, year dummies, personal controls).
There is a smallish but statistically significant degree of auto-regression in the mover equations
of table 10. The coefficient on the lagged dependent variable enters with a coefficient of 0.1336. More
interestingly, the degree of home-ownership in the state has a substantial effect. Its coefficient is −0.8125
with a t-statistic greater than 5. As we add longer and longer lags of home-ownership, going from column
1 to column 6 in table 10, the coefficient drops in size (to a still-substantial −0.3445) but retains its
negative sign and its statistical significance at conventional cut-off levels.
A classic issue in the economics of migration is to what extent workers move long distances. It is
possible to study this within the continental United States by using data on state-to-state moves. Table 11
presents regression results. It takes as its dependent variable the proportion of out-of-state movers, which
is a subset of the movers examined in table 9. Some of the underlying changes in residence may be of a
comparatively short distance—if for example a New Hampshire worker chooses to relocate just over the
state border in Vermont—but on average they will be larger moves than for the within-state data of table 10.
In table 11 our concern is again with whether there is evidence that having a high home-ownership
rate in the state is inimical to mobility. Column 1 of table 11 estimates a mover-rate equation in which
there is a lagged dependent variable, a state unemployment variable, a state union density variable, the
home-ownership variable, a set of year and state dummies, and variables for the degree of education and
personal characteristics of citizens in the state. In this and later columns, the lagged dependent variable
enters with a well-determined coefficient of approximately 0.2. The coefficients on the unemployment
and union variables are not statistically different from zero. Home-ownership, however, enters in a
statistically significant way in column 1 of table 11. At more than unity, its elasticity is large. That number
implies that a doubling of the home-ownership rate would be associated with a halving of the mobility
rate. Column 2 examines the same regression equation but with a one-year lag on the home-ownership
variable. The result is the same and the elasticity now larger. Going to longer lags, however, pushes down
both the coefficient and the degree of statistical significance.
It is possible that the links between high home-ownership and later high unemployment are nothing
to do with the degree of labor mobility. If so, what other processes might be at work? To try to probe
possible mechanisms, table 12 examines whether there is a connection between home-ownership levels
in an area and the ease with which individuals can get to their workplace. Any model with a neoclassical
flavor would suggest that the cost of travelling to work should act as an impediment to the rate of
employment (because it raises the opportunity cost of a job). Table 12 shows that high home-ownership
11
is associated with longer commuting times, which is consistent with the idea that moving for an owneroccupier is expensive, and that in consequence the places with high home-ownership will see more
workers staying put physically but working further from their family home. Because roads, in particular,
are semi-public goods in which individuals can create congestion problems for others, this pattern in the
data is consistent with the existence of un-priced externalities. The elasticity in the final column of table
12 is approximately 0.12.
Table 13 and table 14 turn to the possible concern that—for zoning or NIMBY or other reasons—a
high degree of home-ownership in an area might be associated with a lower degree of tolerance for new
businesses. Table 13 estimates regressions equations in which the dependent variable is the number of
registered firms in the state. State home-ownership enters negatively in these equations with a coefficient
of approximately −0.03 and a t-statistic greater than 2. The right-hand segment of table 13 reports the
same equations for small firms. Once again, the size of these estimated effects in the regression equations
seem to be economically and not just statistically significant. The implied long-run elasticities from
home-ownership on to business formation are large.
Similar patterns emerge using data on US establishments rather than on firms. The results are given
in the state-panel equations of table 14, where the lagged dependent variable has a large coefficient,
and the logarithm of the home-ownership rate in the state enters with a coefficient of approximately
−0.03. Although more research will be required on this topic, and the detailed mechanisms are currently
unknown, these findings are consistent with the view that high home-ownership levels may be inimical to
business formation rates.
CONCLUSIONS
The results in this paper are consistent with the view that high home-ownership impairs the vitality of
the labor market and slowly grinds out greater rates of joblessness. Given the emphasis that most western
governments put on the promotion of home-ownership (one exception is Switzerland, which taxes
homeowners’ imputed rents), it is important that other researchers check and probe these results. Taken
at face value, our findings should be worrying for policy-makers. A possible reason why these patterns
have attracted so little notice from both researchers and the public is that the time lags are long. High
levels of home-ownership do not destroy jobs this year; they tend to do so, on our estimates, the year
after next. Unless these long linkages are studied, the consequences of high levels of home-ownership are
not easy to see.
What mechanism lies behind the findings? It is not possible to be certain. Our main contribution
should be seen as a statistical one—as documenting patterns of potential interest to economists and
social scientists, and perhaps especially to labor economists, macroeconomists, economic geographers,
12
and urban economists. Nevertheless, we have made an attempt to look below the reduced-form link
between current home-ownership and later joblessness. In doing so, we have found evidence that high
home-ownership in a US state is associated with
 lower labor mobility,
 longer commutes, and
 fewer new firms and establishments.
It should be emphasized that this is after we have controlled for state fixed-effects and a range of
possible confounding variables. Our results are consistent with the recent conclusions of a European study
done independently by Laamanen (2013). His study has a number of stronger methodological features
than were available to us.
We have estimated equations using micro data from the United States from the 1980s to the present
day. There are four main conclusions. First, rises in a US state’s home-ownership rate are associated with
later rises in joblessness in that state. The long-run elasticity is estimated to lie between 1 and 2. This
is strikingly large. It suggests that a doubling of home-ownership in a state would be associated in the
steady state with more than a doubling of the unemployment rate. Second, after controlling for state
fixed-effects, we show that areas with higher ownership have lower mobility. The long-run elasticity is
approximately −0.3. Third, high home-ownership areas have longer commute-to-work times, which
can be expected in those areas to raise costs for employers and employees. The long-run elasticity is
approximately −0.1. Fourth, high home-ownership areas have lower rates of business formation. It is
conceivable–we are not able to offer proof—that this may be due to zoning or NIMBY effects. That
conjecture deserves scrutiny in future research.
It is important to emphasize that our study does not claim that homeowners are unemployed more
than renters (very probably they are not). Nor is it an attempt to build on the idea that homeowners are
less mobile than renters (though they probably are). Instead, because the estimates in tables 7 and 9 of
the paper control for whether individuals are themselves renters or owners, the patterns documented in
the paper are consistent with the possibility that the housing market can generate important negative
externalities upon the labor market.
Our analysis has a number of important weaknesses. Unlike Laamanen (2013), we are unable to
assess the effect of exogenous changes in the structure of the housing market. We have had to rely, instead,
upon an examination of the lagged pattern of unemployment observed a number of years after a movement
in a state’s rate of home-ownership. Thus our study adopts the so-called ‘prospective study’ format that
is common in medical science and epidemiology. This is potentially a weakness and means that some
underlying omitted variable, or causal force, might be responsible for the link between Ht and Ut+n. That
13
would not make the patterns in this paper uninteresting ones, but it would mean that a key variable is
missing from the analysis. Another lacuna in our study is a detailed account of the processes by which the
housing market affects the equilibrium rates of unemployment and employment. It may be that the effect
of high home-ownership comes partly from some engendered reduction in the rate of labor mobility within
a geographical area. However, we are doubtful that this works in a classical way through a lower amount of
state-to-state migration.13 Finally, unlike McCormick (1983), we have been unable to distinguish between
those owners who are currently paying a mortgage and those who own their home outright.
Economists currently lack a full understanding of the interplay between the housing and labor
markets. We believe these issues demand the profession’s attention.
REFERENCES
Battu, H., Ma, A., and Phimister, E. 2008. Housing tenure, job mobility and unemployment in the UK.
Economic Journal 118: 311–328.
Bell, D., and Blanchflower, D.G. 2010. UK unemployment in the great recession. National Institute Economic
Review October. London.
Blanchard, O., and Katz, L.F. 1992. Regional evolutions. Brookings Papers on Economic Activity 1: 1–77.
Blanchflower, D.G., and Oswald, A.J. 1994. The wage curve. Cambridge, MA: MIT Press.
Coulson, N.E., and Fisher, L.M. 2002. Tenure choice and labor market outcomes. Housing Studies 17: 35–49.
Coulson, N.E., and Fisher, L.M. 2009. Housing tenure and labor market impacts: The search goes on. Journal
of Urban Economics 65: 252–264.
Dickens, W.T., and Triest, R.K. 2012. Potential effects of the Great Recession on the US labor market. BE
Journal of Macroeconomics 12: article number 8.
DiTella, R., MacCulloch, R.J., and Oswald, A.J. 2003. The macroeconomics of happiness. Review of Economics
and Statistics 85: 809–827.
Dohmen, T.J. 2005. Housing, mobility and unemployment. Regional Science and Urban Economics 35:
305–325.
Dujardin, C., and Goffette-Nagot, F. 2009. Does public housing occupancy increase unemployment? Journal
of Economic Geography 9: 823–851.
Ferreira, F., Gyourko, J., and Tracy J. 2010. Housing busts and household mobility. Journal of Urban Economics
68: 34–45.
Fischel, W.A. 2004. An economic history of zoning and a cure for its exclusionary effects. Urban Studies 41:
317–340.
13. There is evidence—using a variety of statistical methods—that Adam Smith’s compensating differentials theory of
equal utility across geographical space successfully fits the data for the US states (see Roback 1982, for example; or recently
Herz and Van Rens 2012, on the inability of mismatch to explain recent US experience; or in a different way the work
of Oswald and Wu 2010). Consistent with this, new work by Modestino and Dennett (2013) finds little evidence that
house-lock contributes to the pattern of US unemployment.
14
Green, R.K., and Hendershott, P.H. 2001. Home-ownership and unemployment in the US. Urban Studies 38:
1509–1520.
Head, A., and Lloyd-Ellis, H. 2012. Housing liquidity, mobility, and the labor market. Review of Economic
Studies 79: 1559–1589.
Herz, B., and Van Rens, T. 2012. Accounting for mismatch unemployment. Working paper, University of
Warwick. Coventry, England.
Hughes, G., and McCormick, B. 1981. Do council housing policies reduce migration between regions?
Economic Journal 91: 919–937.
Laamanen, J-P. 2013. Home-ownership and the labour market: Evidence from rental housing market
deregulation. Tampere Economic Working Paper, No. 89, University of Tampere, Finland. Available at http://
tampub.uta.fi/handle/10024/68116.
Layard, R., Nickell, S.J., and Jackman, R. 1991. Unemployment: Macroeconomic performance and the labor
market. Oxford University Press: Oxford.
Linn, M.W., Sandifer, R., and Stein, S. 1985. Effects of unemployment on mental and physical health.
American Journal of Public Health 75: 502–506.
McCormick, B. 1983. Housing and unemployment in Great Britain. Oxford Economic Papers 35(S): 283–305.
Modestino, A.S., and Dennett, J. 2013. Are American homeowners locked into their houses? The impact of
housing conditions on state-to-state migration. Regional Science and Urban Economics 43: 322–337.
Mortensen, D.T., and Pissarides, C.A. 1994. Job creation and job destruction in the theory of unemployment.
Review of Economic Studies 61: 397–415.
Munch, J.R., Rosholm, M., and Svarer, M. 2006. Are home owners really more unemployed? Economic Journal
116: 991–1013.
Murphy, G.C., and Athanasou, J.A. 1999. The effect of unemployment on mental health. Journal of
Occupational and Organizational Psychology 72: 83–99.
OECD (Organization for Economic Cooperation and Development). 1994. The Jobs Study. Paris.
Oswald, A.J. 1996. A conjecture on the explanation for high unemployment in the industrialized nations: Part 1.
University of Warwick Economic Research Paper 475.
Oswald, A.J. 1997. Thoughts on NAIRU. Journal of Economic Perspectives 11: 227–228.
Oswald, A.J. 1999. The housing market and Europe’s unemployment: A non-technical paper. Unpublished.
University of Warwick. Coventry, England.
Oswald, A.J., and Wu, S. 2010. Objective confirmation of subjective measures of human well-being: Evidence
from the USA. Science 327: 576–579.
Paul, K.I., and Moser, K. 2009. Unemployment impairs mental health: Meta-analyses. Journal of Vocational
Behavior 74: 264–282.
Petrongolo, B., and Pissarides, C.A. 2001. Looking into the black box: A survey of the matching function.
Journal of Economic Literature 39: 390–431.
Pigou, A. 1914. Unemployment. Holt: New York.
Powdthavee, N. 2010. The happiness equation. Icon Books: Duxford.
Roback, J. 1982. Wages, rents, and the quality of life. Journal of Political Economy 90: 1257–1278.
15
Rouwendal, J., and Nijkamp, P. 2010. Home-ownership and labor market behaviour: Interpreting the
evidence, Environment and Planning A. 42: 419–433.
Serafinelli, M. 2012. Good firms, worker flows, and productivity. Working paper, UC Berkeley.
Smith, T.E., and Zenou, Y. 2003. Spatial mismatch, search effort, and urban spatial structure, Journal of Urban
Economics 54: 129–156.
Valletta, R. 2012. House-lock and structural unemployment. Working paper 2012-25, Federal Reserve Bank of
San Francisco.
Van Leuvensteijn, M., and Koning, P. 2004. The effect of home-ownership on labor mobility in the
Netherlands. Journal of Urban Economics 55: 580–596.
Zabel, J.E. 2012. Migration, housing market, and labor market responses to employment shocks. Journal of
Urban Economics 72: 267–284.
16
Figure 1
Unemployment rates and home ownership across 28 EU and OECD countries
and Switzerland
unemployment rate
30
25
20
y = 0.1704x – 2.5307
R² = 0.1968
15
10
5
0
35
40
45
50
55
60
65
70
75
80
85
90
95
100
home ownership rate
Source: Organization for Economic Cooperation and Development (OECD).
Figure 2 The states of the USA: Changes over half a century 50-year home ownership rate change
1950–2000 versus 60-year unemployment rate change 1950–2010
unemployment rate change
9
8
7
y = 0.1144x + 2.8889
R² = 0.1078
6
5
4
3
2
1
0
-4
-2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
home ownership rate change
-1
-2
Source: Census Bureau.
17
18
62.3
73.0
69.6
68.8
68.0
67.2
Minnesota
Mississippi
Missouri
Montana
Nebraska
Maryland
72.1
67.5
Maine
Michigan
71.3
Louisiana
Massachusetts
68.7
67.2
Kentucky
72.1
Indiana
67.8
69.9
Illinois
Kansas
67.5
Idaho
Iowa
57.7
69.9
Hawaii
67.4
DC
65.7
42.0
Delaware
Georgia
72.1
Connecticut
Florida
65.5
67.5
Colorado
67.0
55.9
California
Arizona
Arkansas
63.1
66.0
Alaska
65.1
69.7
Alabama
2010
67.4
69.1
70.3
72.3
74.6
73.8
61.7
67.7
71.6
67.9
70.8
69.2
72.3
71.4
67.3
72.4
56.5
67.5
70.1
40.8
72.3
66.8
67.3
56.9
69.4
68.0
62.5
72.5
66.2
2000
66.5
67.3
68.8
71.5
71.8
71.0
59.3
65.0
70.5
65.9
69.6
67.9
70.0
70.2
64.2
70.1
53.9
64.9
67.2
38.9
70.2
65.6
62.2
55.6
69.6
64.2
56.1
70.5
64.2
1990
68.4
68.6
69.6
71.0
71.7
72.7
57.5
62.0
70.9
65.5
70.0
70.2
71.8
71.7
62.6
72.0
51.7
65.0
68.3
35.5
69.1
63.9
64.5
55.9
70.5
68.3
58.3
70.1
64.4
1980
66.4
65.7
67.2
66.3
71.5
74.4
57.5
58.8
70.1
63.1
66.9
69.1
71.7
71.7
59.4
70.1
46.9
61.1
68.6
28.2
68.0
62.5
63.4
54.9
66.7
65.3
50.3
66.7
62.9
1970
64.8
64.0
64.3
57.7
72.1
74.4
55.9
64.5
66.5
59.0
64.3
68.9
69.1
71.1
57.8
70.5
41.1
56.2
67.5
30.0
66.9
61.9
63.8
58.4
61.4
63.9
48.3
59.7
61.9
1960
60.6
60.3
57.7
47.8
66.4
67.5
47.9
56.3
62.8
50.3
58.7
63.9
63.4
65.5
50.1
65.5
33.0
46.5
57.6
32.3
58.9
51.1
58.1
54.3
54.5
56.4
54.5
49.4
55.0
1950
47.1
52.0
44.3
33.3
55.2
55.4
38.1
47.4
57.3
36.9
48.0
51.0
51.5
53.1
40.3
57.9
n.a.
30.8
43.6
29.9
47.1
40.5
46.3
43.4
39.7
47.9
n.a.
33.6
43.6
1940
Historical home-ownership rates in the United States: By decade from 2010–1900 (percent)
United States
Table 1
54.3
54.5
49.9
32.5
58.9
59.0
43.5
55.2
61.7
35.0
51.3
56.0
54.7
57.3
46.5
57.0
n.a.
30.6
42.0
38.6
52.1
44.5
50.7
46.1
40.1
44.8
n.a.
34.2
47.8
1930
57.4
60.5
49.5
34.0
60.7
58.9
34.8
49.9
59.6
33.7
51.6
56.9
58.1
54.8
43.8
60.9
n.a.
30.9
42.5
30.3
44.7
37.6
51.6
43.7
45.1
42.8
n.a.
35.0
45.6
1920
59.1
60.0
51.1
34.0
61.9
61.7
33.1
44.0
62.5
32.2
51.6
59.1
58.4
54.8
44.1
68.1
n.a.
30.5
44.2
25.2
40.7
37.3
51.5
49.5
46.6
49.2
n.a.
35.1
45.9
1910
(continues)
56.8
56.6
50.9
34.5
63.5
62.3
35.0
40.0
64.8
31.4
51.5
59.1
60.5
56.1
45.0
71.6
n.a.
30.6
46.8
24.0
36.3
39.0
46.6
46.3
47.7
57.5
n.a.
34.4
46.5
1900
19
66.7
North Carolina
69.3
68.1
68.2
63.7
70.4
70.7
67.2
63.9
73.4
68.1
69.2
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Source: US Census Bureau.
n.a. = not available
69.6
60.7
Rhode Island
Oregon
Pennsylvania
67.2
62.2
Oklahoma
65.4
53.3
New York
67.6
68.5
New Mexico
Ohio
65.4
New Jersey
North Dakota
58.8
71.0
New Hampshire
2010
70.0
68.4
75.2
64.6
68.1
70.6
71.5
63.8
69.9
68.2
72.2
60.0
71.3
64.3
68.4
69.1
66.6
69.4
53.0
70.0
65.6
69.7
60.9
2000
67.8
66.7
74.1
62.6
66.3
69.0
68.1
60.9
68.0
66.1
69.8
59.5
70.6
63.1
68.1
67.5
65.6
68.0
52.2
67.4
64.9
68.2
54.8
1990
69.2
68.2
73.6
65.6
65.6
68.7
70.7
64.3
68.6
69.3
70.2
58.8
69.9
65.1
70.7
68.4
68.7
68.4
48.6
68.1
62.0
67.6
59.6
1980
66.4
69.1
68.9
66.8
62.0
69.1
69.3
64.7
66.7
69.6
66.1
57.9
68.8
66.1
69.2
67.7
68.4
65.4
47.3
66.4
60.9
68.2
58.5
1970
62.2
68.6
64.3
68.5
61.3
66.0
71.7
64.8
63.7
67.2
57.3
54.5
68.3
69.3
67.0
67.4
68.4
60.1
44.8
65.3
61.3
65.1
56.3
1960
54.0
63.5
55.0
65.0
55.1
61.3
65.3
56.7
56.5
62.2
45.1
45.3
59.7
65.3
60.0
61.1
66.2
53.3
37.9
58.8
53.1
58.1
48.7
1950
48.6
54.4
43.7
57.0
48.9
55.9
61.1
42.8
44.1
45.0
30.6
37.4
45.9
55.4
42.8
50.0
49.8
42.4
30.3
57.3
39.4
51.7
46.1
1940
48.3
63.2
45.9
59.4
52.4
59.8
60.9
41.7
46.2
53.1
30.9
41.2
54.4
59.1
41.3
54.4
58.6
44.5
37.1
57.4
48.4
55.0
47.1
1930
Historical home-ownership rates in the United States: By decade from 2010–1900 (percent) (continued)
Nevada
Table 1
51.9
63.6
46.8
54.7
51.1
57.5
60.0
42.8
47.7
61.5
32.2
31.1
45.2
54.8
45.5
51.6
65.3
47.4
30.7
59.4
38.3
49.8
47.6
1920
54.5
64.6
49.5
57.3
51.5
58.5
64.8
45.1
47.0
68.2
30.8
28.3
41.6
60.1
45.4
51.3
75.7
47.3
31.0
70.6
35.0
51.2
53.4
1910
55.2
66.4
54.6
54.5
48.8
60.4
67.8
46.5
46.3
71.2
30.6
28.6
41.2
58.7
54.2
52.5
80.0
46.6
33.2
68.5
34.3
53.9
66.2
1900
20
73.4
68.1
76.5
69.9
59.9
77.2
76.1
75.2
74.2
70.2
70.2
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
75.2
Indiana
69.3
74.9
Illinois
Kansas
67.9
Idaho
Iowa
55.2
70.5
Hawaii
68.4
DC
69.8
41.9
Delaware
Georgia
72.0
Connecticut
Florida
68.3
70.0
Colorado
68.9
57.1
California
Arizona
Arkansas
66.4
68.0
Alaska
67.4
73.2
Alabama
2000
70.1
68.3
74.0
74.5
76.1
77.1
60.6
70.7
75.5
67.1
73.9
70.4
76.6
75.3
69.4
71.7
55.5
70.1
69.2
42.7
75.4
71.8
68.5
58.2
71.2
68.1
65.3
73.2
67.8
2001
68.5
69.4
74.8
74.9
77.3
76.0
62.6
72.0
74.0
67.4
73.7
70.3
73.9
75.1
70.1
73.0
57.9
71.8
68.7
44.1
75.6
71.5
68.9
57.7
70.3
65.6
67.1
73.7
67.9
2002
69.5
71.5
74.0
73.4
77.2
75.6
64.3
71.6
73.7
67.5
74.4
70.3
73.4
74.4
70.7
74.4
58.3
71.4
69.5
43.0
77.2
73.0
71.3
58.9
69.6
67.0
70.0
76.2
68.3
2003
71.2
72.4
72.4
74.0
76.4
77.1
63.8
72.1
74.7
70.6
73.3
69.9
73.2
75.8
72.7
73.7
60.6
70.9
72.2
45.6
77.3
71.7
71.1
59.7
69.1
68.7
67.2
78.0
69.0
2004
70.2
70.4
72.3
78.8
76.5
76.4
63.4
71.2
73.9
72.5
71.6
69.5
73.9
75.0
70.9
74.2
59.8
67.9
72.4
45.8
75.8
70.5
71.0
59.7
69.2
71.1
66.0
76.6
68.9
2005
67.6
69.5
71.9
76.2
75.6
77.4
65.2
72.6
75.3
71.3
71.7
70.0
74.0
74.2
70.4
75.1
59.9
68.5
72.4
45.9
76.8
71.1
70.1
60.2
70.8
71.6
67.2
74.2
68.8
2006
Recent annual home-ownership rates in the United States, 2000–10 (percent)
United States
Table 2a
68.2
67.3
70.4
74.0
73.5
76.4
64.3
71.7
74.3
71.5
72.9
69.4
73.7
73.8
69.4
74.5
60.1
67.6
71.8
47.2
76.8
70.3
70.2
58.3
69.5
70.4
66.6
73.3
68.1
2007
69.6
70.3
71.4
75.4
73.1
75.9
65.7
70.6
73.9
73.5
72.8
68.8
74.0
74.4
68.9
75.0
59.1
68.2
71.1
44.1
76.2
70.7
69.0
57.5
68.9
69.1
66.4
73.0
67.8
2008
70.2
70.7
72.0
75.5
72.9
74.5
65.1
69.6
74.0
71.9
71.2
67.4
72.4
72.0
69.1
75.5
59.5
67.4
70.9
44.9
76.5
70.5
68.4
57.0
68.5
68.9
66.8
74.1
67.4
2009
(continues)
70.4
68.1
71.2
74.8
72.6
74.5
65.3
68.9
73.8
70.4
70.3
67.4
71.1
71.2
68.8
72.4
56.1
67.1
69.3
45.6
74.7
70.8
68.5
56.1
67.9
66.6
65.7
73.2
66.9
2010
21
71.1
70.7
71.3
72.7
65.3
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
70.9
63.8
72.7
68.7
73.9
63.6
75.9
71.8
71.0
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Source: Current Population Survey.
76.5
71.2
South Carolina
74.7
53.4
New York
61.5
73.7
New Mexico
Rhode Island
66.2
New Jersey
Pennsylvania
64.0
69.2
New Hampshire
2000
73.5
72.3
76.4
66.4
75.1
69.8
72.4
63.9
69.7
71.5
76.1
60.1
74.3
65.8
71.5
71.2
71.0
71.3
53.9
70.8
66.5
68.4
64.6
2001
73.0
72.2
77.2
66.9
74.4
70.3
72.8
63.4
70.3
71.5
77.5
59.4
74.0
66.2
69.6
72.1
69.4
70.0
54.8
70.0
66.9
69.5
65.3
2002
72.9
72.8
78.1
65.9
75.0
71.4
73.4
64.5
70.8
70.9
75.0
59.9
73.7
68.0
69.1
72.8
68.7
70.0
54.3
70.3
66.9
74.4
64.8
2003
72.8
73.3
80.3
66.0
73.4
72.0
74.9
65.5
71.6
68.5
76.2
61.5
74.9
69.0
71.1
73.1
70.0
69.8
54.8
71.5
68.8
73.3
65.7
2004
72.8
71.1
81.3
67.6
71.2
74.2
73.9
65.9
72.4
68.4
73.9
63.1
73.3
68.2
72.9
73.3
68.5
70.9
55.9
71.4
70.1
74.0
63.4
2005
73.7
70.2
78.4
66.7
71.1
74.0
73.5
66.0
71.3
70.6
74.2
64.6
73.2
68.1
71.6
72.1
68.3
70.2
55.7
72.0
69.0
74.2
65.7
2006
73.2
70.5
77.6
66.8
71.5
73.7
74.9
66.0
70.2
70.4
74.1
64.9
72.9
65.7
70.3
71.4
66.0
70.3
55.9
71.5
68.3
73.8
63.3
2007
Recent annual home-ownership rates in the United States, 2000–10 (percent) (continued)
Nevada
Table 2a
73.3
70.4
77.8
66.2
70.6
72.8
76.2
65.5
71.7
70.4
73.9
64.5
72.6
66.2
70.4
70.8
66.6
69.4
55.0
70.4
67.3
75.0
63.6
2008
73.8
70.4
78.7
65.5
69.7
74.3
74.1
65.4
71.1
69.6
74.4
62.9
72.2
68.2
69.6
69.7
65.7
70.1
54.4
69.1
65.9
76.0
62.4
2009
73.4
71.0
79.0
64.4
68.7
73.6
72.5
65.3
71.0
70.6
74.8
62.8
72.2
66.3
69.2
69.7
67.1
69.5
54.5
68.6
66.5
74.9
59.7
2010
22
61.7
72.6
72.3
69.5
66.4
69.3
Minnesota
Mississippi
Missouri
Montana
Nebraska
Maryland
72.7
67.8
Maine
Michigan
74.1
Louisiana
Massachusetts
70.2
70.1
Kentucky
71.3
Indiana
72.7
69.9
Illinois
Kansas
62.4
Idaho
Iowa
50.7
69.7
Hawaii
66.5
DC
63.6
37.3
Delaware
Georgia
70.4
Connecticut
Florida
64.7
67.8
Colorado
65.9
53.7
California
Arizona
Arkansas
57.6
65.2
Alaska
64.5
73.7
Alabama
1984
68.5
66.5
69.2
69.6
70.0
70.7
60.5
65.6
73.7
70.2
68.5
68.3
69.9
67.6
60.6
71.0
51.0
62.7
67.2
37.4
70.3
69.0
63.6
54.2
66.6
64.7
61.2
70.4
63.9
1985
68.3
64.4
67.8
70.4
68.0
70.9
60.3
62.8
74.0
70.4
68.1
66.4
69.2
67.6
60.9
69.8
50.9
62.4
66.5
34.6
71.0
68.1
63.7
53.8
67.5
62.5
61.5
70.3
63.8
1986
66.8
65.0
66.1
72.5
68.9
71.7
60.6
62.7
73.2
71.0
67.6
67.9
67.7
69.1
61.0
71.6
50.7
63.9
66.3
35.8
71.1
67.0
61.8
54.3
68.1
63.3
59.7
67.9
64.0
1987
66.6
65.4
64.8
73.7
69.1
72.5
60.0
63.5
72.2
68.5
65.4
68.6
68.3
68.3
61.4
71.5
53.2
64.8
64.9
37.5
70.1
66.5
60.1
54.4
67.0
66.1
57.0
66.5
63.8
1988
67.2
67.9
63.7
72.2
68.3
73.2
58.9
65.5
73.6
66.3
64.9
68.1
69.6
68.2
61.9
70.2
54.7
64.7
64.4
38.7
68.7
66.4
58.6
53.6
66.3
63.9
58.7
67.6
63.9
1989
67.3
69.1
64.0
69.4
68.0
72.3
58.6
64.9
74.2
67.8
65.8
69.0
70.7
67.0
63.0
69.4
55.5
64.3
65.1
36.4
67.7
67.9
59.0
53.8
67.8
64.5
58.4
68.4
63.9
1990
67.5
69.6
64.2
71.8
68.9
70.6
60.2
63.8
72.0
68.9
67.2
69.7
68.4
66.1
63.0
68.4
55.2
65.7
66.1
35.1
70.2
65.5
59.8
54.5
68.6
66.3
57.1
69.9
64.1
1991
68.4
69.9
65.2
70.4
66.7
70.6
61.8
64.8
72.0
66.7
69.0
69.8
66.3
67.6
62.4
70.3
53.8
66.9
66.0
35.0
73.8
66.1
60.9
55.3
70.3
69.3
55.5
70.3
64.1
1992
67.7
69.7
66.4
69.7
65.8
72.3
60.7
65.5
71.9
65.4
68.8
68.9
68.2
68.7
61.8
72.1
52.8
66.5
65.5
35.7
74.1
64.5
61.8
56.0
70.5
69.1
55.4
70.2
64.0
1993
68.0
68.8
68.4
69.2
68.9
72.0
60.6
64.1
72.6
65.8
70.6
69.0
70.1
68.4
64.2
70.7
52.3
63.4
65.7
37.8
70.5
63.8
62.9
55.5
68.1
67.7
58.8
68.5
64.0
1994
67.1
68.7
69.4
71.1
73.3
72.2
60.2
65.8
76.7
65.3
71.2
67.5
71.4
71.0
66.4
72.0
50.2
66.6
66.6
39.2
71.7
68.2
64.6
55.4
67.2
62.9
60.9
70.1
64.7
1995
Annual homeownership rates in the 1980s and 1990s in the United States, 1984–99 (percent)
United States
Table 2b
66.8
68.6
70.2
73.0
75.4
73.3
61.7
66.9
76.5
64.9
73.2
67.5
72.8
74.2
68.2
71.4
50.6
69.3
67.1
40.4
71.5
69.0
64.5
55.0
66.6
62.0
62.9
71.0
65.4
1996
66.7
67.5
70.5
73.7
75.4
73.3
62.3
70.5
74.9
66.4
75.0
66.5
72.7
74.1
68.1
72.3
50.2
70.9
66.9
42.5
69.2
68.1
64.1
55.7
66.7
63.0
67.2
71.3
65.7
1997
69.9
68.6
70.7
75.1
75.4
74.4
61.3
68.7
74.6
66.6
75.1
66.7
72.1
72.6
68.0
72.6
52.8
71.2
66.9
40.3
71.0
69.3
65.2
56.0
66.7
64.3
66.3
72.9
66.3
1998
(continues)
70.9
70.6
72.9
74.9
76.1
76.5
60.3
69.6
77.4
66.8
73.9
67.5
73.9
72.9
67.1
70.3
56.6
71.3
67.6
40.0
71.6
69.1
68.1
55.7
65.6
66.3
66.4
74.8
66.8
1999
23
68.8
North Carolina
69.1
69.6
67.6
62.5
69.9
66.9
68.3
65.7
72.0
65.2
68.8
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Source: Current Population Survey.
71.1
60.9
Rhode Island
Oregon
Pennsylvania
71.0
61.9
Oklahoma
70.1
51.1
New York
67.7
68.0
New Mexico
Ohio
63.4
New Jersey
North Dakota
58.9
67.1
New Hampshire
1984
73.2
63.8
75.9
66.8
68.5
69.5
71.5
60.5
67.6
67.6
72.0
61.4
71.6
61.5
70.5
67.9
69.9
68.0
50.3
68.2
62.3
65.5
57.0
1985
72.0
66.5
76.4
65.1
68.2
69.8
68.0
61.0
67.4
65.9
70.3
62.2
72.3
63.9
69.7
68.2
69.2
68.2
51.3
67.8
63.3
64.8
54.5
1986
68.9
68.2
72.5
64.4
69.0
70.5
69.0
61.1
67.2
66.8
72.8
60.4
71.8
64.6
70.9
68.6
68.9
68.4
52.0
67.2
64.0
66.4
54.1
1987
67.8
68.0
73.2
64.2
69.8
68.7
70.2
59.9
66.9
66.4
73.8
62.0
72.1
64.0
72.1
69.6
67.7
68.3
50.7
65.4
64.8
67.9
54.3
1988
69.6
69.3
74.8
64.2
70.2
69.7
70.4
61.0
67.3
65.8
71.0
61.2
72.8
63.4
71.4
69.6
67.1
69.4
52.3
65.5
65.7
67.0
54.3
1989
68.9
68.3
72.0
61.8
69.8
72.6
70.1
59.7
68.3
66.2
71.4
58.5
73.8
64.4
70.3
68.7
67.2
69.0
53.3
68.6
65.0
65.0
55.8
1990
68.7
68.9
72.4
61.8
68.9
70.8
70.7
59.0
68.0
66.1
73.1
58.2
74.0
65.2
69.2
68.7
65.4
69.3
52.6
69.5
64.8
66.8
55.8
1991
67.9
69.4
73.3
62.5
67.8
70.8
70.0
58.3
67.4
66.5
71.0
56.8
73.1
64.3
68.9
69.1
63.7
68.6
53.3
70.5
64.6
66.6
55.1
1992
67.1
65.7
73.3
63.1
68.5
68.5
68.9
58.7
64.1
65.6
71.1
57.6
72.0
63.8
70.3
68.5
62.7
68.8
52.8
69.1
64.5
65.4
55.8
1993
65.8
64.2
73.7
62.4
69.3
69.4
69.3
59.7
65.2
66.4
72.0
56.5
71.8
63.9
68.5
67.4
63.3
68.7
52.5
66.8
64.1
65.1
55.8
1994
69.0
67.5
73.1
61.6
68.1
70.4
71.5
61.4
67.0
67.5
71.3
57.9
71.5
63.2
69.8
67.9
67.3
70.1
52.7
67.0
64.9
66.0
58.6
1995
68.0
68.2
74.3
63.1
68.5
70.3
72.7
61.8
68.8
67.8
72.9
56.6
71.7
63.1
68.4
69.2
68.2
70.4
52.7
67.1
64.6
65.0
61.1
1996
Annual homeownership rates in the 1980s and 1990s in the United States, 1984–99 (percent) (continued)
Nevada
Table 2b
67.6
68.3
74.6
62.9
68.4
69.1
72.5
61.5
70.2
67.6
74.1
58.7
73.3
61.0
68.5
69.0
68.1
70.2
52.6
69.6
63.1
66.8
61.2
1997
70.0
70.1
74.8
64.9
69.4
69.1
73.7
62.5
71.3
67.3
76.6
59.8
73.9
63.4
69.7
70.7
68.0
71.3
52.8
71.3
63.1
69.6
61.4
1998
69.8
70.9
74.8
64.8
71.2
69.1
74.7
62.9
71.9
70.7
77.1
60.6
75.2
64.3
71.5
70.7
70.1
71.7
52.8
72.6
64.5
70.2
63.7
1999
Table 3
Unemployment equations without a lagged dependent variable—estimated using
state-year cell means
1985–2011
Log home-ownership ratet-1
1986–2011
1987–2011
1988–2011
1989–2011
–.3282 (2.09)
Log home-ownership ratet-2
1989–2011
–.8309 (3.39)
.0031 (0.02)
.2540 (0.78)
Log home-ownership ratet-3
.2083 (1.29)
Log home-ownership ratet-4
.0012 (0.00)
.4588 (2.87)
Log home-ownership ratet-5
–.1520 (0.48)
.8060 (5.18)
1.0216 (4.40)
Union density
–.4232 (1.08)
–.5939 (1.47)
–.7233 (0.47)
–.8038 (1.99)
–.1693 (0.70)
–.6592 (1.66)
Year dummies
25
24
23
22
21
21
State dummies
50
50
50
50
50
50
Education dummies
18
18
18
18
18
18
4
4
4
4
4
4
Personal controls
N
1377
1326
1275
1224
1173
1173
Adjusted R2
.7831
.7792
0.7822
.7958
.8151
.8175
Notes: The dependent variable in this table is the log of the state unemployment rate in year t. The personal controls here are age, gender, 15 level-ofeducation variables, and two race dummies. t-statistics are in parentheses.
Source: Calculated from the Merged Outgoing Rotation Group (MORG) files of the US Current Population Survey, 1985–2011.
Table 4
Unemployment equations with a lagged dependent variable—estimated using state-year
cell means
Log unemployment ratet-1
Log home-ownership ratet-1
1985–2011
1986–2011
1987–2011
1988–2011
1989–2011
1989–2011
.8482 (50.67)
.8536 (50.86)
.8442 (51.37)
.8173 (49.08)
.7840 (45.77)
.7860 (45.24)
.2488 (2.73)
Log home-ownership ratet-2
–.1460 (1.01)
.3359 (3.69)
.3303 (1.73)
Log home-ownership ratet-3
.2927 (3.26)
–.0837 (0.44)
Log home-ownership ratet-4
.3429 (3.79)
Log home-ownership ratet-5
.4302 (4.47)
.4081 (2.97)
–.0171 (0.09)
Union density
–.1619 (0.71)
–.1041 (0.61)
–.1066 (0.47)
–.1109 (0.48)
–.1693 (0.70)
–.1402 (0.60)
Year dummies
25
24
23
22
21
21
State dummies
50
50
50
50
50
50
Education dummies
18
18
18
18
18
18
4
4
4
4
4
4
Personal controls
N
1377
1326
1275
1224
1173
1173
Adjusted R2
.9283
.9292
.9330
.9349
.9323
.9371
Notes: The dependent variable in this table is the log of the state unemployment rate in year t. The personal controls here are age, gender, 15 level-ofeducation variables, and two race dummies. t-statistics are in parentheses. If the equation of column 1 is re-estimated with contemporaneous homeownership then home-ownership has a t-statistic less than 2; the exact result for the right hand side of that equation is as follows (with t-statistics in
parentheses):
0.8447 (50.52) Logunt-1 + 0.1154 (1.26) log home – 0.1407 (0.64) union density.
Source: Calculated from the Merged Outgoing Rotation Group (MORG) files of the US Current Population Survey, 1985–2011.
24
Figure 3 The impulse response function in log unemployment
after a one unit rise in homeownership
2.5
2
1.5
1
0.5
Home
In
0
-10
0
10
20
30
40
Note: This uses the representative equation: lnun(t) = .784*lnun(t–1) + .4302 home(t–5)
Table 5
Evidence of robustness across two sub-periods: Unemployment equations with
a lagged dependent variable
1989–2001
Log unemployment ratet-1
.7214 (28.18)
Log home-ownership ratet-1
–.0814 (0.44)
2002–2011
.7169 (28.23)
.6839 (18.31)
.6939 (19.05)
.6884 (18.87)
–.2113 (0.81)
Log home-ownership ratet-2
.3862 (1.59)
.0344 (0.11)
Log home-ownership ratet-3
.1081 (0.44)
–.1486 (0.50)
Log home-ownership ratet-4
–.3381 (1.45)
Log home-ownership ratet-5
.5162 (2.81)
.3566 (2.66)
.3042 (1.24)
Union density
–.5202 (1.48)
–.4778 (1.37)
.3853 (1.24)
Year dummies
12
12
9
9
9
State dummies
50
50
50
50
50
Education dummies
18
18
15
15
15
4
4
4
4
4
663
663
510
510
510
0.9307
.9296
.9439
.9440
.9391
Personal controls
N
Adjusted R
2
.6098 (2.07)
.6867 (3.64)
.6246 (3.26)
.3343 (0.75)
Note: The dependent variable is the log of the state unemployment rate in year t. t-statistics are in parentheses.
Source: MORG files of the CPS.
25
3
.3236 (0.72)
Table 6
Employment equations with a lagged dependent variable—estimated using state-year
cell means
1985–2011
1986–2011
1987–2011
1988–2011
1989–2011
1989–2011
Log employment ratet-1
.8846 (64.67)
.8837 (62.66)
.8897 (62.44)
.8771 (60.01)
.8581 (56.29)
.8576 (56.10)
Log home-ownership ratet-1
–.0164 (1.72)
Log home-ownership ratet-2
.0259 (1.66)
–.0317 (3.30)
Log home-ownership ratet-3
–.0330 (1.59)
–.0298 (3.10)
–.0013 (0.07)
Log home-ownership ratet-4
–.0313 (3.21)
Log home-ownership ratet-5
–.0431 (4.32)
Union density
.0337 (1.42)
.0468 (1.94)
.0414 (1.72)
Year dummies
25
24
23
22
.0107 (0.54)
–.0457 (3.07)
.0414 (1.65)
.0414 (1.63)
21
21
State dummies
50
50
50
50
50
50
Education dummies
18
18
18
18
18
18
4
4
4
4
4
4
Personal controls
N
1377
1326
1275
1224
1173
1173
Adjusted R2
.9832
.9831
.9838
.9841
.9842
.9842
Notes: The dependent variable in this table is the log of the state employment rate, in year t. t-statistics are in parentheses. The personal controls here are
age, gender, 15 education variables, and two race dummies.
Source: MORG files of the US Current Population Survey, 1985–2011.
26
Table 7 Weeks-worked equations 1992–2011—estimated using micro data
All
Log home-ownership ratet-4
–.7679 (2.14)
Log home-ownership ratet-5
Home owner
Non-movers
–.7123 (1.95)
–.6960 (1.94)
–.6740 (1.81)
.7529 (10.77)
.7527 (10.76)
.6190 (8.75)
.6192 (8.75)
.3734 (5.26)
.3731 (5.26)
.2569 (3.50)
.2573 (3.50)
Non-mover
.6593 (24.44)
.6592 (24.44)
n.a.
n.a.
Union density
1.8295 (2.08)
1.7019 (1.94)
1.2324 (1.40)
1.1125 (1.27)
3.7152
3.4327
4.2496
4.1088
Year dummies
25
24
23
22
State dummies
50
50
50
50
Education dummies
15
15
15
15
7
7
7
7
N
2,268,197
2,268,197
1,981,950
1,981,950
R2
.8075
.8075
.8216
.8216
Renter
Constant
Personal controls
n.a. = not available
Note: The dependent variable here is the number of weeks that that individual worked in the previous year. Personal
controls include gender, race and a full-time variable. Excluded category: rent in lieu of cash. Standard errors clustered
at the state/year cell. t-statistics are in parentheses.
Source: March CPS.
27
Table 8
US data on the annual rate of mobility (defined as moving
residence) and the home-ownership rate
Percent
movers
Same
county
2010–11
11.6
2009–10
2008–09
Homeownership
rate
Same state
Different
state
Abroad
7.7
2.0
1.6
0.4
66.9
12.5
8.7
2.1
1.4
0.3
67.4
12.5
8.4
2.1
1.6
0.4
67.8
2007–08
11.9
7.8
2.1
1.6
0.4
68.2
2006–07
13.2
8.6
2.5
1.7
0.4
68.8
2005–06
13.7
8.6
2.8
2.0
0.4
68.9
2004–05
13.9
7.9
2.7
2.6
0.6
69.0
2003–04
13.7
7.9
2.8
2.6
0.4
68.3
2002–03
14.2
8.3
2.7
2.7
0.4
67.9
2001–02
14.8
8.5
2.9
2.8
0.6
67.8
2000–01
14.2
8.0
2.7
2.8
0.6
67.4
2011
66.2
1999–2000
16.1
9.0
3.3
3.1
0.6
66.8
1998–99
15.9
9.4
3.1
2.8
0.5
66.3
1997–98
16.0
10.2
3.0
2.4
0.5
65.7
1996–97
16.5
10.5
3.0
2.4
0.5
65.4
1995–96
16.3
10.3
3.1
2.5
0.5
64.8
1994–95
16.4
10.8
3.1
2.2
0.3
64.0
1993–94
16.7
10.4
3.2
2.6
0.5
64.0
1992–93
17.0
10.7
3.1
2.7
0.6
64.2
1991–92
17.3
10.7
3.2
2.9
0.5
64.1
1990–91
17.0
10.3
3.2
2.9
0.6
64.0
1989–90
17.9
10.6
3.3
3.3
0.6
63.9
1988–89
17.8
10.9
3.3
3.0
0.6
63.8
1987–88
17.8
11.0
3.3
3.0
0.5
64.0
1986–87
18.6
11.6
3.7
2.8
0.5
63.8
1985–86
18.6
11.3
3.7
3.0
0.5
63.9
1984–85
20.2
13.1
3.5
3.0
0.6
64.5
1983–84
17.3
10.4
3.6
2.8
0.5
64.7
1982–83
16.6
10.1
3.3
2.7
0.4
64.8
1981–82
17.0
10.3
3.3
3.0
0.5
65.4
1980–81
17.2
10.4
3.4
2.8
0.6
65.6
17.7
10.8
3.4
3.0
0.6
1970–71
18.7
11.4
3.1
3.4
0.8
64.2
1969–70
19.1
11.7
3.1
3.6
0.8
64.3
1968–69
19
11.7
3.2
3.4
0.7
63.9
1977–79
1975–76
64.9
1972–75
64.7
64.5
(continues)
15
28
Table 8
US data on the annual rate of mobility (defined as moving
residence) and the home-ownership rate (continued)
Same state
Different
state
Abroad
Homeownership
rate
11.8
3.4
3.6
0.7
63.6
11.6
3.3
3.4
0.7
63.5
19.8
12.7
3.3
3.3
0.5
63.0
20.7
13.4
3.5
3.3
0.5
1963–64
20.1
13.0
3.3
3.3
0.5
1962–63
20.0
12.6
3.1
3.6
0.6
1961–62
19.6
13.0
3.0
3.1
0.5
1960–61
20.6
13.7
3.1
3.2
0.6
1959–60
19.9
12.9
3.3
3.2
0.5
1958–59
19.7
13.1
3.2
3.0
0.5
1957–58
20.3
13.1
3.4
3.3
0.5
1956–57
19.9
13.1
3.2
3.1
0.5
1955–56
21.1
13.7
3.6
3.1
0.6
1954–55
20.4
13.3
3.5
3.1
0.6
1953–54
19.3
12.2
3.2
3.2
0.6
1952–53
20.6
13.5
3.0
3.6
0.5
1951–52
20.3
13.2
3.2
3.4
0.4
1950–51
21.2
13.9
3.6
3.5
0.2
1949–50
19.1
13.1
3.0
2.6
0.3
1948–49
19.2
13.0
2.8
3.0
0.3
1947–48
20.2
13.6
3.3
3.1
0.3
Percent
movers
Same
county
1967–68
19.5
1966–67
19.0
1965–66
1964–65
Notes: For four different definitions of mobility—within the county, within the state, across states, and overseas
Source: Census Bureau (available at http://www.census.gov/hhes/migration/data/cps/cps2011.html).
29
16
Table 9
Mover equations where the dependent variable is whether the individual
moved residence, 1989–2010
Log unemployment rate
Log home-ownership rate
.0127 (3.44)
.0116 (3.33)
.0116 (2.57)
.0079 (2.54)
.0079 (1.78)
–.1766 (9.20)
–.1542 (8.44)
–.1542 (4.09)
–.0633 (3.85)
–.0633 (1.62)
Renter
.2134 (73.45)
.2134 (18.19)
Renter (in lieu of cash)
.1144 (38.52)
.1144 (27.18)
19
19
19
Year dummies
22
State dummies
50
50
50
50
50
0
21
21
21
21
State & year
State & year
State
State & year
State
N
2,941,082
2,941,082
2,941,082
2,941,082
2,941,082
R2
.0109
0.491
.0491
.1208
.1208
Personal controls
Clustered standard errors
19
Source: March CPS. The mover variable relates to a question asked in March of that year asking whether the respondent had moved
over the preceding year. So the home-ownership and unemployment variables relate to the previous year, 2010. The personal
home-ownership variables and other personal controls relate to the date of interview i.e., March 2011. Personal controls here are
age, gender, four race dummies, and 15 level-of-education dummies. t-statistics are in parentheses.
Table 10
Moving rate equations—estimated using state-year cell means
Log mover ratet-1
Log unemployment ratet-1
.1336 (4.06)
.1506 (4.63)
.1639 (5.03)
.1724 (5.32)
.1742 (5.39)
.1700 (5.12)
.0419 (1.64)
.0468 (1.83)
.0557 (2.17)
.0605 (2.36)
.0629 (2.45)
.0451 (1.67)
Log union density ratet-1
–.0494 (1.29)
–.0460 (1.19)
–.0564 (1.45)
–.0628 (1.62)
–.0662 (1.71)
–.0667 (1.67)
Log home-ownership ratet
–.8125 (5.45)
Log home-ownership ratet-1
–.6950 (4.70)
Log home-ownership ratet-2
–.4446 (3.00)
Log home-ownership ratet-3
–.3242 (2.23)
Log home-ownership ratet-4
–.3085 (2.15)
Log home-ownership ratet-5
–.3445 (2.36)
Year dummies
20
20
20
20
20
19
Education dummies
15
15
15
15
15
15
State dummies
50
50
50
50
50
50
5
5
5
5
5
5
1071
1071
1071
1071
1071
1020
.8424
.8424
.8403
.8397
.8396
.8396
Personal control variables
N
R
2
Note: Here the dependent variable is the log of the proportion of movers per annum within the state. As before, the data are answers given to whether
the respondent had moved over the preceding year. So the home-ownership and unemployment variables relate to the previous year, 2010. The personal
home-ownership variables and other personal controls relate to the date of interview i.e., March 2011. Personal controls are age, gender, four race dummies,
and 15 education dummies. t-statistics are in parentheses.
Source: March CPS.
30
Table 11
Moving rate equations for out-of-state movers—estimated using state-year cell means
Log mover ratet-1
Log unemployment ratet-1
Log union density ratet-1
Log home-ownership ratet
.2155 (6.84)
.2155 (6.87)
.2187 (6.91)
.2206 (7.02)
.2212 (7.03)
.2117(6.48)
–.0317 (0.45)
–.0339 (0.48)
–.0125 (0.18)
–.0047 (0.07)
.0016 (0.02)
–.0051 (0.07)
.0352 (1.29)
.0559 (0.52)
.0276 (0.33)
.0251 (0.24)
.0103 (0.12)
.0184 (0.17)
–1.0396 (2.57)
Log home-ownership ratet-1
-1.3104 (3.25)
Log home-ownership ratet-2
–.7302 (1.80)
Log home-ownership ratet -3
–.7977 (2.00)
Log home-ownership ratet-4
–.6513 (1.65)
Log home-ownership ratet-5
–.5736 (1.43)
Year dummies
20
20
20
20
20
19
Education dummies
15
15
15
15
15
15
State dummies
50
50
50
50
50
50
5
5
5
5
5
5
1071
1071
1071
1071
1071
1020
.7276
.7277
.7267
.7269
.7265
.7258
Personal control variables
N
R
2
Note: Here the dependent variable is the log of the proportion of movers per annum who left the state. The personal controls relate to the date of
interview i.e., March 2011. Personal controls are age, gender, four race dummies and 15 education dummies. Here the mover variable is as before but is
now defined as the proportion that moved out of state in the state-year cell. t-statistics are in parentheses.
Source: March CPS.
Table 12
Log commuting-time equations, 2005–10 (in minutes)—estimated
using state-year cell means
Log home-ownership ratet
Year dummies
–.6208
–.6294
–.3533
.1145
.1196
(14.60)
(14.30)
(5.47)
(2.35)
(2.42)
—
5
—
5
5
State dummies
—
—
50
50
50
Education dummies
—
—
—
—
15
Personal controls
—
—
—
—
10
.0040
.0042
.0208
.0209
.0415
7,509,307
7,509,307
7,509,307
7,509,307
7,509,307
N
Adjusted R2
N
Note: Here the dependent variable is the log of the average commuting time per annum in the state. Personal controls
are age and its square, male, and seven race dummies. Standard errors are clustered by state and year. t-statistics are
in parentheses. The mean of commuting time is 25.2 minutes (one way); it has a standard deviation of 22.2 minutes.
Source: American Community Surveys Micro Data files.
31
Table 13
Log number-of-firms equations, 1988–2010—estimated using state-year
cell means
Log number of firmst
Log number of firmst-1
.9417
(112.57)
.9416
(113.87)
Log number of small firmst
.9416
(106.55)
Log number of small firmst-1
Log home-ownership ratet-4
.9497
(115.56)
–.0297
(2.78)
Log home-ownership ratet-5
.9504
(119.12)
.9504
(118.65)
–.0408
(2.68)
–.0408
(1.95)
–.0278
(2.14)
–.0334
(3.01)
–.0333
(2.21)
Year dummies
23
22
22
23
22
22
State dummies
51
51
51
51
51
51
Education dummies
15
15
15
15
15
15
Personal controls
10
10
10
10
10
10
Standard errors clustered by
N
2
R
State & year
State & year
State
State & year
State & year
State
1122
1122
1122
1122
1122
1122
.9999
.9999
.9999
.9999
.9999
.9999
N
Note: Here the dependent variable is the log of the number of firms per annum in the state. Personal controls are age and its square, male
and seven race dummies. t-statistics are in parentheses.
Source: Calculated from the Business Dynamics files. See http://www.census.gov//econ/susb/data/susb2010.html and http://www.sba.
gov/advocacy/849/12162#susb.
32
Table 14
Number-of-establishments equations,
1988–2010—estimated using state-year
cell means
Log number of establishmentst
Log number of establishmentst-1
Log home-ownership ratet-4
.9414
.9411
.9411
(113.60)
(115.14)
(109.07)
–.0317
–.0317
(3.16)
(2.36)
–.0300
(2.96)
Log home-ownership ratet-5
Year dummies
23
22
22
State dummies
51
51
51
Education dummies
15
15
15
Personal controls
10
10
10
Standard errors clustered by:
State & year
N
2
R
State & year
1122
1122
.9999
.9999
State
.9999
Notes: Here the dependent variable is the log of the number of establishments per
annum in the state. Personal controls are age and its square, male and seven race
dummies. Standard errors are clustered by state and year. t-statistics are in parentheses.
Source: Calculated from the Business Dynamics files. See http://www.census.gov//econ/
susb/data/susb2010.html and http://www.sba.gov/advocacy/849/12162#susb.
33
Appendix Table A1
Log unemployment ratet-1
Checking for an effect from unemployment benefits (the replacement rate):
Log unemployment rate equations estimated using state year cells using the
Merged Outgoing Rotation Group (MORG) files of the Current Population
Survey, 1989–2011
1989–2000
2001–2011
.7855 (45.70)
.7871 (45.88)
.7843 (45.76)
.7877 (45.81)
.7370 (28.07)
.6664 (21.67)
.4315 (4.73)
.4241 (4.65)
.4291 (4.70)
.4252 (4.66)
.3279 (2.29)
.5953 (3.33)
Union density
–.1120 (0.48)
–.1112 (0.48)
–.1165 (0.50)
–.1088 (0.47)
–.4438 (1.24)
.2237 (0.54)
U.I. replacement ratet
–.0006 (1.04)
–.0327 (2.05)
–.0274 (1.41)
–.0444 (1.71)
Log home-ownership ratet-5
U.I. replacement ratet-1
–.0004 (0.60)
–.0347 (2.22)
U.I. replacement ratet-2
–.0110 (0.69)
Year dummies
21
21
21
21
11
10
State dummies
50
50
50
50
50
50
Education dummies
18
18
18
18
18
15
4
4
4
4
4
4
1173
1173
1275
1224
612
561
.9371
.9373
.9330
.9349
.9375
.9371
Personal controls
N
2
Adjusted R
Notes: The personal controls are age, gender, and two race dummies. Here the replacement rate is defined as average weekly benefit amount defined
below divided by average weekly earnings from the MORGs. t-statistics are in parentheses.
Source: Of UI data—see http://workforcesecurity.doleta.gov/unemploy/hb394.asp from US Dept of Labor Employment and Training Administration
ET Financial Data HANDBOOK 394 Report FEDERAL-STATE EXTENDED BENEFITS REPORT FOR 1983 THROUGH 2012 UNDER PUBLIC LAW 91-373.
34